Crypto Alpha: Where Top Allocators Are Finding Opportunities in This Bull Market

In a discussion about finding alpha in the current, cautiously optimistic bull market, three prominent allocators introduced their investment strategies. Arthur Cheong, founder and CEO of Defiance Capital, described his firm as a crypto liquid fund that has been investing since 2020, focusing on a classic liquid venture style by purchasing crypto assets with promising potential.

Rajiv Patel-O'Connor, a General Partner at Framework Ventures, an SF-based investment firm, specializes in early-stage venture and some liquid token investing. Smac, a partner at Compound, manages a thesis-driven, research-centric investment firm based in New York. Compound typically engages in early-stage venture but has recently expanded its focus to public equities and tokens, with approximately one-third of the fund dedicated to crypto-related assets.

Arthur Cheong observed a significant shift in the crypto-AI landscape since 2023. Initially, efforts to integrate crypto and AI often involved 'force-fitting' themes, which proved challenging. However, over the past year, more coherent and sensible applications have emerged.

Cheong highlighted Hermes, a major open-source AI agent framework, as an example of a legitimate AI team that originated from crypto-AI development. Additionally, Venise AI, described as a crypto-native version of OpenRouter for privacy-focused AI inference, has shown substantial growth. While its token spend and inference volume are not yet comparable to OpenRouter, its growth chart indicates rapid adoption, suggesting increasing legitimacy of crypto infrastructure within broader AI development.

Rajiv Patel-O'Connor noted that AI is fundamentally changing how companies operate by providing operational leverage, enabling startups to make significant progress with smaller teams and reduced headcount expenses. This transformation is particularly evident in the financing opportunities within the crypto sector.

Patel-O'Connor highlighted crypto's effectiveness in capital formation, especially for the extensive AI capex buildout. He cited USDA, a portfolio company, as an example, which offers mid-market cluster financing for GPU acquisition. This model can provide promising startups and public companies seeking to purchase large volumes of GPUs, such as $150 million worth, with potentially more attractive rates and lower down payments compared to traditional financing options.

Investing in AI builders requires evaluating both talent and underlying societal values. While crypto-AI currently faces challenges competing solely on efficiency with centralized models, its unique value proposition lies elsewhere. Key differentiators for crypto-AI include permissionless systems, decentralized coordination, and innovative economic models.

Arthur Cheong emphasized that crypto-AI projects typically incorporate encryption and privacy by default. This contrasts sharply with many centralized AI projects, which often prioritize rapid development over privacy considerations. Cheong believes that as the world increasingly values privacy, the solutions already present in crypto-AI projects will gain significant traction.

Smac added that crypto's historical strengths in coordination and economic incentives are particularly relevant. He also highlighted the global and permissionless nature of crypto as a crucial advantage, offering a hedge against growing concerns about data ownership and privacy, especially given the rising pushback against centralized frontier AI models.

Smac expressed that crypto has historically excelled at coordination and economic incentives. As blockchains become more efficient, the argument for crypto in AI agents revolves around facilitating machine-to-machine transactions and communication. While centralized systems can enable agents to communicate using credit cards, there is a minimum cost floor for these transactions. The question remains whether crypto AI can reduce these costs to a level that centralized providers cannot match, though this is still to be determined.

Beyond cost, the global and permissionless nature of crypto is a significant factor. Concerns about data ownership and privacy are growing, and there is a definite trend of pushback against traditional, centralized AI models and frontier labs. Smac believes that crypto offers opportunities within these areas, providing a hedge against centralized control.

The discussion touched on the idea that crypto was not initially built for human utilization, prompting the question of whether its use by AI agents in the future is a retrospective 'force-fitting' or a genuine evolution. Smac's perspective centers on crypto's core strengths, which align well with the needs of AI agents for efficient, coordinated, and economically incentivized interactions.

Smac suggested that if the future of AI involves more local or smaller models rather than exclusively massive frontier models, with inference increasingly handled by distributed versions, then this aligns well with the core beliefs of crypto. These beliefs center around distributed networks and privacy. However, the question of value capture for crypto projects within this shift remains open.

The sector is still in its nascent stages, making it challenging to forecast definitive outcomes. Despite the uncertainty surrounding value capture, the directional alignment between distributed AI inference and crypto's foundational principles suggests a potential synergy.

Felix noted that top crypto funds are currently adopting a 'barbell' allocation strategy, effectively balancing illiquid venture-stage investments with active participation in liquid markets. This approach has proven effective for successful crypto funds that have endured and continue to allocate capital. The privacy narrative, in particular, has gained significant momentum recently, presenting interesting challenges and opportunities.

Many privacy-focused assets, such as Zcash and zero-knowledge proof tokens, are publicly tradable. However, building substantial positions in these liquid assets can be difficult compared to the venture book, where sizing into investments is often easier and liquidity concerns are managed differently, usually only becoming critical at the point of a liquidity event.

Arthur Cheong argues that building significant positions in crypto protocols is not a major issue for investors with high conviction. He observed that individuals have been able to accumulate a few percent of any given protocol over a long period. The difficulty of execution primarily arises for funds managing over $500 million to $1 billion in assets.

Cheong also noted that many investors historically preferred venture-style investing to avoid the technical complexities of liquid market execution. However, he emphasized that venture investing still involves underwriting the volatility of outcomes, even if daily price movements are not constantly monitored. For liquid allocations, investors must be willing to overlook day-to-day volatility while building positions. His firm typically budgets for a 5-10% variance in final entry and exit prices, aiming for a 5% variance, suggesting that a willingness to accept this level of slippage makes building positions manageable.

Rajiv Patel-O'Connor stated that Framework Ventures seeks venture-style outcomes for all its investments, whether in liquid assets or private markets. This approach often necessitates a willingness to hold an asset over a very long time horizon to realize its volatility and potential convexity.

Sizing into assets can be challenging, particularly with limited liquidity or if teams are reluctant to engage in OTC deals. The market is also increasingly crowded, with ideas becoming consensus more quickly, leading to rapid price movements as other market participants conduct similar research. To manage this, Patel-O'Connor emphasizes understanding the upside case and being comfortable with some variance in entry price.

He also explained that Framework Ventures, historically focused on private venture seed and pre-seed, expanded into the liquid side to gain broader asset class exposure when they found fewer early-stage founders meeting their investment criteria. This shift reflects a strategic adaptation to market conditions while maintaining a long-term, venture-oriented perspective.

Felix observed that the current bull cycle exhibits significantly more dispersion compared to previous cycles, where asset prices tended to move in unison. This dispersion creates opportunities for active managers to generate alpha, distinct from merely riding levered beta. However, this environment also leads to rapid crowding out, as market participants quickly converge on a small number of attractive assets, making it challenging to establish positions.

Rajiv Patel-O'Connor responded that managing this dispersion involves careful consideration of timing and time horizons. Assets that appear highly attractive can quickly become consensus and crowded in the short term, but may be less so over longer periods. Managers are thus adjusting their entry strategies and investment horizons to differentiate their positions from the rapidly forming consensus, recognizing that this cycle feels fundamentally different and requires continuous learning.

Arthur Cheong stated that the era of easy returns in crypto, where simply buying any asset led to gains, is ending as the market matures and sophistication increases. He believes that getting into assets early and sizing them appropriately is becoming more difficult. Unlike public equities, crypto cycles unfold much faster, often within 3 to 6 months, necessitating quick decisions on entry and exit points.

This increased dispersion in the market is beneficial for active managers, who leverage it to demonstrate their value proposition to Limited Partners (LPs). Cheong noted that, year-to-date, over 80% of liquid crypto hedge funds known to him have outperformed Bitcoin and ETH. These funds are reporting returns ranging from 20% to 100%, indicating a healthy phenomenon where good fundamentals are being rewarded more swiftly.

Arthur Cheong observed that even highly consensus assets like $HYPE, with significant market caps, can continue to outperform if they possess strong fundamentals and growth prospects. This marks a shift from past cycles where consensus often signaled a peak for an asset.

He also identified a bifurcation in market participation: 'smart money' investors focus on fundamentally strong assets, while retail investors, no longer content with 50-200% annual returns, are seeking 10-20x gains in extreme 'left-curve' assets, specifically low-cap meme coins ranging from $100k to $50 million in market capitalization. This indicates a divergence in risk appetite and return expectations within the crypto market.

Arthur Cheong highlighted a growing consensus among 'smart money' investors who are gravitating towards major and mid-cap assets with solid fundamentals. This disciplined approach aims to consistently outperform Bitcoin, a crucial factor for future capital raising and overall fund health. Cheong expressed a desire for this trend to continue, indicating a healthier and more sophisticated market.

Investors are now compelled to model the marginal buyer for any asset they acquire today. This involves anticipating market dynamics and understanding that the speed at which investment ideas gain traction and compress value is much faster than in the past. This requires a proactive stance, where investors must quickly identify fundamental alignments, take initial positions, and conduct further due diligence over time, rather than passively observing.

Arthur Cheong affirmed that Bitcoin remains the primary benchmark for most Limited Partners (LPs), who still perceive crypto as a single asset class. Despite attempts to create alternative indices like CCI30 or the Bloomberg crypto index, they have not gained significant popularity among allocators. For better or worse, Bitcoin continues to be the most crucial benchmark for LPs, who prioritize both absolute returns and outperformance against Bitcoin.

The rise of Bitcoin ETFs, such as IBIT, has further solidified Bitcoin's role as a baseline. Fund managers are now expected to at least outperform Bitcoin; otherwise, LPs might question the value of allocating to a fund instead of directly purchasing an ETF. However, Smac added that sophisticated LPs are also increasingly benchmarking funds systematically against their peers, indicating a more nuanced approach to performance evaluation.

The venture capital landscape within crypto is evolving, moving beyond an earlier obsession with speculative seed-stage funding. The industry is maturing, and there is now greater clarity on which projects have achieved product-market fit and possess key use cases. This shift prompts investors to re-evaluate their allocation strategies between nascent seed-stage potential and established growth-stage entities.

Arthur Cheong observed an extreme overlap among late-stage deals in both equity and crypto. This phenomenon is largely due to a scarcity of late-stage deals, meaning many investors are vying for a limited pool of opportunities. He noted that these late-stage deals typically perform well, making them highly sought after by those who can secure an allocation.

Arthur Cheong noted a significant increase in seed-stage opportunities in crypto, partly because many investors were 'burned' in previous cycles. He emphasized that crypto and AI are unique in their ability to deliver 50x returns within one to two years if the seed stage is executed effectively. He doubts many other industries can offer such rapid, substantial uplifts within a two-year timeframe.

However, securing top-tier founders for seed-stage projects remains challenging. Cheong pointed out that the AI sector is currently attracting many high-quality founders, intensifying the competition for talent in crypto. Rajiv Patel-O'Connor added that while the number of outlier founders has not dramatically decreased, there are fewer seed-stage founders compared to two or three years ago, partly due to a right-sizing of the capital deployed at this stage.

Rajiv Patel-O'Connor explained that his firm maintains a specialized investment strategy, focusing on pre-seed and seed-stage opportunities. They aim to avoid 'scope creep' by not expanding into public markets or later growth-stage private investments, believing their expertise lies specifically in these early stages. They do not see themselves as competitive in the 'publics and everything in between' space.

Patel-O'Connor also expressed a 'naively optimistic' view that some crypto projects previously deemed 'broken' or unsuccessful might find success in the future. He believes that many past attempts were simply ill-timed three, four, or five years ago, and with the right founders and improved timing, these concepts could now thrive.

Rajiv Patel-O'Connor discussed how AI tools are fundamentally altering team structures and capital intensity in early-stage crypto startups. He noted that portfolio companies are now able to operate with fewer employees because team members are becoming 'AI-native,' effectively leveraging these tools. This shift means that employees who are slow to adopt AI tools are no longer considered net-additive to the team.

While headcount costs are decreasing, budgets are being reallocated from salaries to compute resources and API credits. However, certain functions, such as enterprise sales and extensive marketing, still require human interaction. Patel-O'Connor pointed out that a $10 million early-stage raise now extends much further, unless the startup is involved in training foundational AI models, which remains an extremely capital-intensive endeavor that quickly depletes early-stage funding.